Background: The optimal system for lymph node (LN) staging in gastric cancer is still a subject of debate. The aim of our study was to analyse the probability of error in negative LN (pN0) gastric carcinomas when a low number of LNs were harvested using a probabilistic model.
Methods: Patients with gastric carcinoma who underwent R0 resection at three university hospitals between 2004 and 2009 were retrospectively included. A Bayesian model was used to analyse the probability of error for negative LNs (pN0) gastric carcinomas. Kaplan-Meier survival curves and the log-rank test were used to compare the overall and specific mortality of prognosis groups.
Results: Of the 291 patients included, 123 were classified as pN0 (42%). A significant correlation was found between the extent of the LN dissection performed and the number of the LNs retrieved. According to the Bayesian model the carcinomas with 9 or fewer negative lymph nodes were considered to have a high risk (HR) of misclassification, whereas patients with 10-25 LNs analysed and those with more than 26 negative lymph nodes were considered to have a moderate risk (MR) and low risk (LR), respectively. The log-rank test showed a significant improvement in the disease-specific survival for the MR pN0 (p < 0.001) and LR pN0 (p < 0.04) but not for the HR pN0 patients compared to pN1 patients.
Conclusions: The proposed probabilistic model is clinically useful for differentiating the prognosis in pN0 gastric carcinomas when an insufficient number of negative lymph nodes are retrieved.
Keywords: Bayes theorem; Gastric cancer; Lymph nodes; Lymphadenectomy; Probability; Prognosis.
Copyright © 2013 Elsevier Ltd. All rights reserved.